|
SONY DSC-U60: Twinning for 3D Photography John Hart Program in Atmospheric and Oceanic Sciences University of Colorado Boulder, CO 80302 nimbus.colorado.edu/hart/science.htm August 28, 2003 Updated Jan 2, 2005
The Sony DSC-U60 is a small waterproof (to 5 feet) and rugged digital camera. Some sample pictures and comments on its imaging performance are contained in our earlier review. Full specifications are found on the dpreview website (among other places). In this TechNote I explain how to parallel wire the shutter and power buttons on a pair of U60's to obtain a small and highly-synchronized digital stereo camera. It even has an instant viewer.
SPECIFICATIONS OF STEREO-TWIN RIG:
In the following I describe how to disassemble and parallel wire this camera, give synchronization test results, and present some 3D test pairs. Before attempting the re-wiring project I constructed a simultaneous mechanical trigger. I wanted to see how well you could do with respect to synchronization by just pressing down the two shutter button as closely together as mechanically possible. Some people do digital 3D co-firing cameras with two fingers. Here is an attempt to see what you can do with a fairly stiff and accurate mechanical coupling.
The sync was measured with the spinning wheel (see below). The results show that on average you can expect to get about 1/40 'th second sync, with an occasional mis-fire. How much better can we do by parallel wiring the buttons themselves, and pressing one (or the other) to trigger the cameras?
CAMERA DISASSEMBLY First you have to get the cameras apart. Some precision tools are handy: binocular magnifier, illuminator, tweezers, the smallest Phillips head screwdriver you can find, a bent pair of tweezer with rubberized tips (a do-it-yourself tool), and a micro-point soldering iron.
The first step is to un-screw the LCD cover (4 screws), with the camera turned off and the batteries removed out the base, of course. Note the rugged case. The goal is to slide the whole camera out the bottom (which opens and is sealed with an O-ring).
After the LCD cover is removed, use the bent tweezers to gently coax the LCD panel out of the square hole. This will reveal its connector. The connector has two tiny little plastic clamps on the sides. Carefully loosen these with a tweezer tip. Now try to grab grab the connector uniformly across the black tape (see pix below), using the bent-tip tweezers, and smoothly and carefully slide the ribbon micro-cable out of its connector. This allows you to take the LCD panel off and out through the case. THIS, AND PUTTING THE LCD PANEL BACK (just reverse the steps) IS THE HARDEST PART OF THE WHOLE SURGICAL PROCEDURE.
Once the LCD panel is off, it's a simple matter to unscrew the three retaining screws in the bottom of the camera housing, and then slide the whole camera assembly out of the case. The two cases, the left one with the camera housing removed, are shown below.
Since I needed to test the thing as I was going along, I put the LCD panel back on the camera body. You don't have do this (the less times fooling with the micro-connector the better). Next, locate the shutter button and the power button.
Below, for reference, is an photograph of the front of the camera and case. It's good to keep a dust free environment. A clean tech-spray dust removal aerosol should work OK. Before you reassemble, make sure there is no dust on the lens or on the inside of the oval glass on the front of the housing.
This is tiny stuff. Pretty remarkable fabrication and assembly (for the money).
Looking down on the shutter button. Here is how to solder it. Get the smallest wire you can find. I used three-wind solid wire, a couple feet long. Solder both buttons in parallel. The tip on the iron I used was 0.03", and the solder 0.02".
Two-wires are all that are needed for the power button. There are 4 terminals on this button, but only the top two are used.
I ran the wires down the sides and out the base. Be careful to position them below the buttons.
Here the wires come out through triangular cut in the base, then make a right angle turn and go through a 1/16 hole in the shell. This would allow you to disassemble (god forbid) the camera without having to cut the wires, even after re-waterproofing the rig as shown below.
The wires come through the side of the housing. Devcon High Strength Plastic Welder epoxy is used to seal the hole and cover the wires as they run to a 6 pin terminal that has been glued to a plexiglass cross-bar. You will probably need the terminal block since it is easiest to wire and seal the cameras individually, then to join the 5 wires at a terminal. The cross bar is epoxied to the camera bases after they are aligned with the upper (larger) plexiglass panel.
The main attachment is made out of plexiglass (1/4 thick). 3-48 screws are installed in the two LCD panel covers so that they point outwards. It is necessary to sand the heads to get them to go into the small slots in the blue LCD-cover plastic, outside the seal. The plexiglass panel is cut and drilled for the 3-48's and in order to access the screws that hold the LCD panel in place (8 total, 4 for each camera). The cameras are held in a gripper (a small vice on a tripod) and pointed towards a scene with horizontal lines (not the images shown in the LCD panels above). After lightly sanding the surfaces that will meet, apply a small amount of Devcon Plastic Welder to the LCD panels. Loosely tighten the bolts. With the cameras turned on adjust the cameras so that there is no rotational error in a distant scene with horizontal lines (before the epoxy hardens, of course). Then tighten the 3-48 nuts. After the upper attachment is dry, the lower cable bar can be flipped over and glued to the base. In all this, don't get glue where you don't want it (e.g. the LCD panels, the base seals, etc.). Wet glue can be removed with ethyl alcohol.
SYNCHRONIZATION TESTS I returned to the spinning wheel to test the units. I didn't feel I needed huge accuracy, so the absolute scales could be off by as much as 10% or so. Individual readings have errors estimated to be plus or minus a thousandth sec or so. Basically, the wheel spins at 4 Hz and each division is about 1/250'th second. I did the tests in the sun, with the cameras set to SPORT mode - which forces a short shutter opening. All shots in these test were made by first half-pressing the right camera shutter button until both the left and right cameras indicated that focus and exposure were preset. Then I fully pressed the right shutter button to fire them. The composite picture, below, is from the two cameras. You can see that the motion of the wheel is pretty well frozen. The left camera leads the right by about 0.3 tick marks, 1/800'th sec or thereabouts!
In a long series of pictures, the sync drifts as the phase of the oscillator (the switch polling) shifts slowly with time. This is exactly how the other digital cameras I tested worked. The only difference is that the Sony U60 cameras have a beautiful little feature: THE POWER ON/OFF IS NEARLY INSTANTANEOUS, WITH NO MECHANICAL MOTIONS (like lenses coming out, etc.). POWER ON IS 0.9 seconds (similar or less for off). Therefore a typical operating scenario would be: 1) Power the cameras on (since the power buttons are parallel wired they both come on together) 2) Hopefully this puts them into sync (see below). 3) Shoot a few pictures and power off. Go to 1).
The chart above shows how the sync errors grow with time after an initial power reset (i.e. power off, then power on). For about 3 things are great then they start to fall apart. Those who read my Nikon CP5000 report will note the similarity in the sync behavior with time. After about 20 minutes (for the Sony cameras) the polling cycle is coming back into phase and the cameras will sync up again (for about 3 minutes), etc., etc. But by using the near-instantaneous reset you can initiate a good sync state easily. Or can you?
This chart (above) shows what happens for about 100 power-resets. We switch the cameras on (press one of the power buttons), set both to Sport Mode, and shoot one pair of images (in VGA resolution). Continue this procedure every 30 seconds until the 8MB memories that came with the cameras are full. There are 9 cases where the sync is worse than 1/250'th (Note: 1/250 sync is an acceptable value for photographing most moving subjects with this wide-angle-lens camera). In 10% of the resets, sync is relatively poor. In 3% of resets the sync is bad (like 1/30'th), though even 1/30'th will work for many slowly moving subjects. Actually the situation is probably better than the data shown above. My previous work with the CP5000 led to the following theory: If the polling cycle is P = 0.04 seconds (for example), and if the sync at the start of a run is S = 0.001 seconds (for example), then, S/P ( = 1/40'th) of the time, the sync will be about 0.04 seconds. Some of the extreme ( ~ 0.04) "flyers" are due to this stochastic effect.
Here is the average sync-error (absolute) of 5 shots spaced 30 seconds apart following a power up reset. The behavior is very good except for one case. In that case the errors were randomly -2/256 and 8/256 seconds. As with the Nikon CP5000, the sum of the errors is ~ 1/25'th (0.04) . I anticipate that this is the switch polling interval for these cameras. In conclusion, it appears that 90% of the time sync will be excellent, provided a power reset is used every few minutes. This is naturally what you will do with these cameras anyway, in order to preserve battery power. AN "IN-THE-FIELD" 3D VIEWER Now here is a nice variation. These cameras are so narrow and perfectly positioned that you can look at the screens and see your stereo image instantly. At first I did this with a pair of prism glasses, which works pretty well if there is no ambient light. Then I thought, why not just take the face plate off a 3D slide viewer and make a little clamp to hold it onto the back of the twin-cam. Yes indeed, instant 3D. Not the highest resolution of course (the camera's LCD panel has about 300 x 240 dots), but enough to gauge the shot! The images below show how it works.
A first version of the Digital Twin-Cam. All the epoxy is in place to water-proof the interconnections. I like the second version's mounting better.
Franka Slide Viewers with and without the plastic slide holder. A spring loaded mounting bracket has been added to the one on the right.
The 3D Viewer-Cam.
Side-View. TEST PICTURES (Some quick target-of-opportunity shots to test the sync). Click any picture from the table view it. Use the BACK button (or BACKSPACE on Internet Explorer) to return to the picture table matrix.
During the field tests, about 30 fast-action shots were made (all above except 1 and 8 fit that category). 4 of the 30 had serious sync issues. This is in accord with the finding in the lab that the twins synced well about 90% of the time. CLICK HERE FOR MORE IMAGE DISCUSSION CONCLUSION: Parallel wiring the shutter/preset and power buttons on a Sony DSC-U60 proved relatively easy to do and led to vastly improved camera synchronization over manual triggering and shutter/preset wiring only. The cameras are intrinsically able to sync to better than 1/500'th second, and do so on roughly 90% of the shots provided power is cycled every few minutes. This naturally occurs as you usually turn these cameras off to preserved battery life between shots. Image quality from these 2MPixel cameras is reasonably good, yielding sharp enough images for the web, for XGA projection, and for stereo-cards from most shots. Fine detail in low contrast areas tends to go soft more often than I would like. On the other hand, the ease of use and rough waterproof nature of the rig will make it possible to obtain stereo-pairs in situations that would otherwise be very difficult or impossible. An interesting little project. Now, off to some wet places to hopefully get some unique stereo imagery of waves and turbulence. |
